Re: IP cameras on ring topology, not star

[Pat Coghlan <info@xxxxxxxxxx>]

T-splice is a good description of what I need to do atop each pole.

I think this should be possible, in that it's kind of a miniature
version of a metropolitan area network (MAN).  In this case, though, I'm
connecting a single host (camera) at each add/drop location on the ring
rather than a "building".

Even if the traffic from each camera had to flow around the ring
independent of the other cameras back to a large switch, that would be
okay.  In other words, all cameras use the ring purely as a medium to
connect them back to a switch.

Matt Ion wrote:
> Pat Coghlan wrote:
>> Matt Ion wrote:
>>
>>> Pat Coghlan wrote:
>>>
>>>> By hub on each pole I mean a device connected to the ring, with an
>>>> ethernet port for 1 camera.
>>>
>>>
>>> Umm, okay... what you're basically talking about then, is a
>>> three-port hub on each pole: one port for the camera, one to uplink
>>> to the previous pole, one to uplink to the next pole.  It's do-able
>>> (4-port 10/100 ethernet switches are fairly cheap these days) but
>>> rather convoluted.
>>
>>
>> I'm not suggesting daisy-chaining up to a couple of hundred of these
>> devices around 2 km ring.
>>
>> Let's get back to the main requirement: cameras must be connected in a
>> ring rather than home-running (to borrow Morgan's description) each
>> camera back to a switch.  I'd like to find out what the best way to do
>> this would be.
>
> That's just the point: I don't think there is one, short of a "faux"
> ring as I desscibed above.  What you're basically looking for is some
> way to "tee-splice" each camera into a single wiring ring... you can't
> do that with video OR network, period.  The closest would be as I
> desrcibed above, using a small switch at each pole as the "tee" connector.

I'm not an optical expert, but a technology like FDDI, Sonet or DWM
could provide the ring.

>
> The only type of network that would PHYSICALLY resemble this would be a
> 10base-2 setup (the old ethernet-over-coax), but then you'd have to find
> IP cameras with a 10base-2 interface (non-existant, as 10base-2 is even
> more obsolete than token-ring), and you'd also be limited to 10
> megabits, which would not be nearly enough bandwidth.
>
> The important question is, WHY *must* they be in a ring, rather than
> home-run?  Perhaps the customer has unrealistic requirements or
> expectations.

The reason is, of course, the fact that if the poles were, say 50m
apart, there would be a 50m run to camera #1, a 100m run to camera #2
etc.  It was suggested that cameras be grouped together, but the team
looking at setting up/tearing down this project essentially want to
drive up with a truck loaded with pre-cut cables and go pole-to-pole.

>> I kinda wondered if there might be adapter hubs that do this,
>> especially whether they buffered traffic etc. until the token arrived.
>
> You're talking about an ethernet to token-ring bridge, something like
> this:
> http://www.ringdale.com/products/st/asp/control.wizmoreinfo/id.227/po./en/default.html
>
>
> However, token-ring wiring is STILL PHYSICALLY done as a home-run.

Yeah, that's appears to be the deal killer with token ring.

>> The advantage of token ring would be that it resolves the collision
>> issue that occurs with CSMACD.
>
> That it does, but when you're talking 100 IP cameras, I don't think
> you'll find it has enough bandwidth.  And again, it doesn't solve the
> desire for a physical ring layout.

Agreed.

>> Cost is not an object.
>
> How about putting a little platform on each pole and just have a
> security guard sit there with a handycam?

We tried that.  Won't fly.

> Again, I think you're looking at a small ethernet switch mounted on each
> pole to act as a "tee" to a daisy-chained pole-to-pole ethernet run.  If
> you're going gigabit, make sure to use AT LEAST Cat-5e cabling (Cat-6
> preferable, since cost is not an object).

Out of curiosity, what's a reasonable limit on how many can be
daisy-chained together?

The optical ring approach might fit the bill.

--

-Pat